Steps in Descriptive Data Analysis Using Python

Descriptive analysis is the process of summarising, organising and presenting data so that researchers can understand the main characteristics of a dataset. In Python, descriptive analysis can be conducted efficiently using libraries such as Pandas, NumPy, SciPy and Matplotlib.

Below is a practical step-by-step procedure suitable for students, researchers, NGOs and organisations.

1. Install the Required Python Libraries

The commonly used libraries are:

pip install pandas numpy scipy matplotlib seaborn openpyxl

The main purposes are:

• Pandas – data management and descriptive statistics
• NumPy – numerical calculations
• SciPy – statistical analysis
• Matplotlib – graphs and charts
• Seaborn – statistical visualisations
• Openpyxl – reading and writing Excel files

2. Import the Libraries

Start your Python analysis by importing the required packages.

import pandas as pd
import numpy as np
import matplotlib.pyplot as plt
import seaborn as sns
from scipy import stats

3. Import Your Dataset

For an Excel dataset:

data = pd.read_excel("research_data.xlsx")

For a CSV dataset:

data = pd.read_csv("research_data.csv")

You can then display the first observations:

data.head()

To display the last observations:

data.tail()

4. Examine the Structure of the Dataset

Before conducting analysis, determine the size and structure of your data.

data.shape

This returns:

(number of rows, number of columns)

For example:

(250, 15)

means that the dataset contains 250 observations and 15 variables.

You can also examine the variable names:

data.columns

5. Check the Data Types

Use:

data.info()

This helps identify whether variables are:

• Integer
• Float
• Text/string
• Date/time
• Boolean

Correct data types are important because statistical calculations depend on the type of variable.

6. Check for Missing Values

Missing data can affect your analysis.

Use:

data.isnull().sum()

To calculate the percentage of missing observations:

data.isnull().mean() * 100

For example:

You can then decide whether to exclude, replace or otherwise handle missing observations based on the research design.

7. Clean the Data

Before calculating descriptive statistics, check for errors and inconsistencies.

For example:

data["Age"].describe()

If you find impossible values such as an age of 250, investigate the original questionnaire or dataset.

You can also remove duplicate observations:

data = data.drop_duplicates()

8. Generate Frequencies for Categorical Variables

Frequency analysis is commonly used for variables such as:

• Gender
• Education level
• Marital status
• Occupation
• Enterprise type
• Location

For example:

data["Gender"].value_counts()

To obtain percentages:

data["Gender"].value_counts(normalize=True) * 100

You can combine frequencies and percentages:

frequency = data["Gender"].value_counts()
percentage = data["Gender"].value_counts(normalize=True) * 100

result = pd.DataFrame({
    "Frequency": frequency,
    "Percentage": percentage
})

print(result)

9. Calculate Measures of Central Tendency

Descriptive analysis normally includes measures of central tendency:

Mean

data["Income"].mean()

Median

data["Income"].median()

Mode

data["Income"].mode()

The mean is the arithmetic average, while the median represents the middle observation when values are ordered. The mode is the most frequently occurring value.

10. Calculate Measures of Dispersion

You can calculate the spread of the observations using:

Minimum

data["Income"].min()

Maximum

data["Income"].max()

Range

data["Income"].max() - data["Income"].min()

Variance

data["Income"].var()

Standard deviation

data["Income"].std()

Standard deviation is particularly useful for understanding how widely observations vary around the mean.

11. Generate a Complete Descriptive Statistics Table

Pandas provides a convenient command:

data.describe()

For numerical variables, this produces:

• Count
• Mean
• Standard deviation
• Minimum
• 25th percentile
• Median
• 75th percentile
• Maximum

For example:

             Age       Income
count       100.0       100.0
mean         38.4     850000.0
std           9.2     320000.0
min          20.0     250000.0
25%          31.0     600000.0
50%          37.0     800000.0
75%          45.0    1000000.0
max          67.0    2500000.0

12. Analyse Categorical Variables

You can generate descriptive information for categorical variables using:

data["Education"].value_counts()

For several categorical variables:

categorical_variables = ["Gender", "Education", "Marital_Status"]

for variable in categorical_variables:
    print("\n", variable)
    print(data[variable].value_counts())

13. Analyse Numerical Variables

Select numerical variables:

numeric_data = data.select_dtypes(include=np.number)

Then:

numeric_data.describe()

This provides a summary of all numerical variables.

14. Examine Relationships Between Variables

Although descriptive analysis primarily summarises individual variables, researchers often examine simple relationships between numerical variables.

For example, correlation:

data["Income"].corr(data["Age"])

For correlations among several variables:

data.corr(numeric_only=True)

A correlation matrix can also be visualised:

plt.figure(figsize=(10, 7))
sns.heatmap(data.corr(numeric_only=True), annot=True)
plt.title("Correlation Matrix")
plt.show()

15. Create Frequency Charts

A bar chart can be used to present categorical data.

data["Gender"].value_counts().plot(kind="bar")

plt.title("Distribution of Respondents by Gender")
plt.xlabel("Gender")
plt.ylabel("Frequency")
plt.show()

16. Create a Histogram

Histograms are useful for examining the distribution of numerical variables.

For example:

plt.hist(data["Age"], bins=10)

plt.title("Distribution of Respondents by Age")
plt.xlabel("Age")
plt.ylabel("Frequency")
plt.show()

17. Create a Pie Chart

For simple categorical distributions:

data["Gender"].value_counts().plot(
    kind="pie",
    autopct="%1.1f%%"
)

plt.title("Respondents by Gender")
plt.ylabel("")
plt.show()

For academic research, however, bar charts are often easier to interpret than pie charts, particularly when there are many categories.

18. Create a Boxplot

A boxplot helps identify the distribution and potential outliers.

sns.boxplot(x=data["Income"])

plt.title("Distribution of Income")
plt.show()

This can help researchers identify unusually high or low observations that require investigation.

19. Cross-Tabulation

Cross-tabulation is useful when examining two categorical variables.

For example, gender and education:

pd.crosstab(
    data["Gender"],
    data["Education"]
)

To display row percentages:

pd.crosstab(
    data["Gender"],
    data["Education"],
    normalize="index"
) * 100

This can be useful in survey research.

20. Export Results to Excel

After producing your descriptive statistics, you can export them to Excel.

descriptive = data.describe()

descriptive.to_excel("descriptive_statistics.xlsx")

You can also export frequency tables:

frequency.to_excel("gender_frequency.xlsx")

21. Interpret the Results

Generating statistics is only one part of research analysis. The researcher must also interpret what the statistics mean.

For example, suppose the analysis produces:

• Mean age = 38.4 years
• Median age = 37 years
• Standard deviation = 9.2 years
• Minimum = 20 years
• Maximum = 67 years

A research report might state:

The findings indicate that respondents had a mean age of 38.4 years, with a standard deviation of 9.2 years. The youngest respondent was 20 years old, while the oldest was 67 years old, indicating variation in the age distribution of respondents.

The interpretation should be based on the actual results rather than simply describing Python commands.

22. Present the Results in a Research Report

A typical descriptive-analysis section may contain:

Table 1: Demographic characteristics

Table 2: Descriptive statistics

The tables can then be followed by a written interpretation explaining the major findings.

Complete Python Example

The following is a simple workflow that can be adapted to a research dataset:

import pandas as pd
import numpy as np
import matplotlib.pyplot as plt
import seaborn as sns

# 1. Import data
data = pd.read_excel("research_data.xlsx")

# 2. Inspect data
print(data.head())
print(data.shape)
print(data.info())

# 3. Check missing values
print(data.isnull().sum())

# 4. Descriptive statistics
print(data.describe())

# 5. Gender frequency
gender_frequency = data["Gender"].value_counts()
gender_percentage = data["Gender"].value_counts(normalize=True) * 100

gender_table = pd.DataFrame({
    "Frequency": gender_frequency,
    "Percentage": gender_percentage
})

print(gender_table)

# 6. Mean, median and standard deviation
print("Mean age:", data["Age"].mean())
print("Median age:", data["Age"].median())
print("Standard deviation:", data["Age"].std())

# 7. Frequency chart
data["Gender"].value_counts().plot(kind="bar")

plt.title("Distribution of Respondents by Gender")
plt.xlabel("Gender")
plt.ylabel("Frequency")
plt.show()

# 8. Histogram
plt.hist(data["Age"], bins=10)

plt.title("Distribution of Respondents by Age")
plt.xlabel("Age")
plt.ylabel("Frequency")
plt.show()

# 9. Correlation matrix
correlation = data.corr(numeric_only=True)
print(correlation)

# 10. Export descriptive statistics
data.describe().to_excel("descriptive_statistics.xlsx")

Recommended Research Workflow

For a student, NGO or organisational research project, the overall Python workflow can therefore be structured as:

Import data → Inspect data → Clean data → Handle missing values → Code variables → Generate frequencies → Calculate percentages → Calculate mean/median/mode → Calculate standard deviation → Examine distributions → Generate cross-tabulations → Create charts → Export tables → Interpret findings → Write the results chapter.

This workflow can be extended from simple descriptive statistics to correlation, t-tests, ANOVA, chi-square tests and regression analysis when these techniques are appropriate for the research objectives and data.